MPLAB XC8 C Compiler Version 1.10 Release...

MPLAB® XC8 C CompilerVersion 1.10 Release Notes

Produced on Wednesday, 1 August 2012

THIS FILE CONTAINS IMPORTANT INFORMATION RELATING TO THIS COMPILER.PLEASE READ IT BEFORE RUNNING THIS SOFTWARE.

Overview 2Introduction 2Previous Versions 2Devices Supported 2Editions and License Upgrades 2Installation and Activation 2Compiler Documentation 3Customer Support 3

Documentation Updates 3New Documents 3Error/Warning Messages 4

What's New 4Version 1.10 4Version 1.01 5Version 1.00 5

Migration Issues 6Version 1.10 6Version 1.01 7Version 1.00 7

Fixed Issues 8Version 1.10 8Version 1.01 9Version 1.00 11

Known Issues 15Microchip Errata 20

MPLAB® XC8 C Compiler Version 1.10 Release Notes" 1

1. Overview1.1. IntroductionThis is a signifiant update to the Microchip MPLAB XC8 C compiler.This compiler introduces a new compiler interface (CCI) to enhance portability between all MPLAB XC C compilers, new part support, fixes a number of reported bugs, adds new optimiza-tions and additional functionality. A number of limitations associated with previous compiler versions have been lifted.The XC8 compiler encompasses the two HI-TECH C compilers which previously supported Mi-crochip 8-bit PIC devices: HI-TECH C Compiler for PIC10/12/16 MCUs and HI-TECH C Com-piler for PIC18 MCUs. This MPLAB XC8 compiler supports compilation for any supported 8-bit PIC device from the one application.

1.2. Previous VersionsThe previous version MPLAB XC8 C compiler was 1.01, released June 2012.The last released version of HI-TECH C Compiler for PIC10/12/16 MCUs was 9.83, released in September 2011. The previous HI-TECH C Compiler for PIC18 MCUs was 9.80, released in Oc-tober 2011.

1.3. Devices SupportedThis compiler supports all known 8-bit PIC devices at the time of release. See pic_chipinfo.html (in the compiler's doc directory) for a list of all supported baseline and mid-range devices and pic18_chipinfo.html for a list of all supported PIC18 devices.

1.4. Editions and License UpgradesThe MPLAB XC8 compiler can be activated in several different operating modes: Free, Standard and PRO. The Standard and PRO modes are licensed modes and require a key for activation ob-tained after purchase. These two modes offer two levels of improved optimization features com-pared to the Free mode. A compiler operating in Free mode can be operated indefinitely without a license. You can change to any operating mode (license permitting) from the MPLAB X IDE project properties (or MPLAB IDE v8 Build Options) on a build-by-build basis.The compiler can be evaluated in any mode for 60 days free of charge. After the evaluation pe-riod has expired, the compiler reverts to the Free mode, and the trial optimization features are disabled.You can choose the operating mode or choose to evaluate the compiler during the installation and activation process.

1.5. Installation and ActivationSee also the Migration Issues and Limitations sections for important information about the latest license manager included with this compiler.


If using MPLAB IDE, be sure to install MPLAB IDE v8.10 (or later) or MPLAB X before in-stalling these tools. Quit the IDE before installing the compiler. Run the .exe (Windows), .run (Linux) or .app (OS X) compiler installer application, e.g. XC8-1.00.11403-windows.exe and follow the directions on the screen. The default installation directory is recommended. If you are using Linux, you must install the compiler from a root account.The option "Update MPLAB IDE to use XC8 for existing C18 projects", if set, will change your MPLAB IDE settings so that all existing legacy projects designated to use either MPLAB C18 or MPLAB C Compiler for PIC18 MCUs, will, by default, use this MPLAB XC8 compiler in C18 compatibility mode. To revert this change if you select it here, you will need to manually adjust the toolsuite settings for legacy projects that should continue to use the MPLAB C18 or MPLAB C Compiler for PIC18 MCUs compilers. This option has no effect if you do not have legacy pro-jects.See the document License Manager for MPLAB® XC C Compilers (DS52059) for more infor-mation.

1.6. Compiler DocumentationThe compiler's user's guide covers all aspects of the compiler's operation, as well as other useful information. Check the well populated index for your search term.

1.7. Customer SupportCommon problems are explained in the FAQ list. You can also ask questions of other users of this product in the XC8 Forum.Microchip welcomes bug reports, suggestions or comments regarding this compiler version. Please direct any bug reports or feature requests via the Support System.At times, advisory message 1395 may be issued by the compiler. This message is part of a new testing process. The compiler will display this message if it encounters a specific code sequence that results in internal compiler templates being used in a unique way. It does not mean there is a bug in the generated code, but that the code sequence encountered could be used to further im-prove the compiler's performance. If you wish to participate by contributing the code that gener-ated this message, you are welcome to send the project to; otherwise, you may ignore this mes-sage.

2. Documentation Updates2.1. New DocumentsThe compiler user's guide has been updated with a new chapters dedicated to the Common C In-terface and a new how-to section. The guide also details new compiler features, and expands the description of some existing features.A new PDF version of the PIC18 peripheral library document has been added to the compiler docs directory. This indicates those functions which are relevant to each device and the library interface.


http://www.microchip.com/stellent/idcplg?IdcService=SS_GET_PAGE&nodeId=2903&param=en549920

http://www.microchip.com/stellent/idcplg?IdcService=SS_GET_PAGE&nodeId=2903&param=en549920

http://www.microchip.com/forums/f249.aspx

http://www.microchip.com/forums/f249.aspx

http://support.microchip.com

http://support.microchip.com

2.2. Error/Warning MessagesThe following are messages that are not present in the compiler's user's guide.1435 variable "*" is incompatible with other objects in section "*" (Code Generator)

You cannot place variables that have differing startup initializations into the same psect. That is, variables that are cleared at startup and variables that are assigned an initial non-zero value must be in different psects. Similarly, bit objects cannot be mixed with byte objects, like char or int.int __section("myData") input; // okay

int __section("myData") output; // okay

int __section("myData") level = 0x1234; // oops -- not in myData

bit __section("myData") mode; // oops again -- no bits with bytes

// each different object to their own new section

1436 "*" is not a valid nibble, use hexadecimal digits only (Parser)When using __IDLOC(), the argument must only consist of hexadecimal digits with no radix specifiers or other characters. Any character which is not a hexadecimal digit will be programmed as a 0 in the corresponding location.__IDLOC(0x51); // oops -- you cannot use the 0x radix modifier

3. What's NewThe following are new features the compiler now supports. The version number in the subhead-ings indicates the first compiler version to support the features that follow.

3.1. Version 1.10New device support The following parts are now fully supported by this release: MCP19110,

MCP19111, RF675F, RF675H, RF675K, PIC12F529T39A, PIC12F529T48A, PIC12LF1840T39A, PIC12LF1552, PIC16F527, PIC16F1454, PIC16LF1454, PIC16F1455, PIC16LF1455, PIC16F1459, PIC16LF1459, PIC16F1784, PIC16LF1784, PIC16F1786, PIC16LF1786, PIC16F1787, PIC16LF1787, PIC18F45K50, PIC18F24K50, PIC18F25K50, PIC18LF45K50, PIC18LF24K50, PIC18LF25K50, PIC18F97J94, PIC18F87J94, PIC18F67J94, PIC18F96J94, PIC18F86J94, PIC18F66J94, PIC18F95J94, PIC18F65J94, PIC18F85J94, PIC18F96J99, PIC18F86J99, PIC18F66J99.

The Common C Interface (CCI) The Common C Interface is a documented set of ANSI stan-dard refinements, non-standard extensions and guidelines to help you write code which is more portable across all MPLAB XC C compilers. A new chapter has been added to the XC8 User's Guide describing these features.

User's guide A new compiler user's guide has been included with this release. See the Documen-tation Updates section, above, for more information.

Roam-out Licensing A new “roam out” feature allows a network license to be checked out for an extended period for use on a particular computer. While the license is checked out, the


computer has licensed use to an XC compiler, and need not be in contact with the network license server. When the license is returned to the network license server, it is once more available to be used as a floating license, or to be roamed out to other computers.

Psect allocation The CCI __section() specifier can also be used in non-CCI mode. Refer to the CCI chapter in the user's guide. It can be used in place of the #pragma psect direc-tive.

Function and module information Information about each function, which appears in the as-sembly list file, is now also displayed in the map file. In addition, a summary of program memory usage on a module-by-module basis is shown in the map file. This allows an esti-mate of the size of the code (excluding data) being generated by each module.

Bank specification with PIC18 devices The qualifiers bank0 through bank3 may now be used with PIC18 devices to allocate variables to a specific memory bank. These qualifiers must be used with the --ADDRQUAL option.

Implementation of strftime function The strftime() function has been implemented and is available in the standard libraries.

Qualifier synonyms A side effect of the CCI features is that when not in CCI or strict ANSI mode, most of the non-standard specifiers, e.g. bit, near and far, can also be represented with two leading underscores, as in __bit, __near and __far etc.

SFR structure types The structures that are defined to represent SFRs are now a typedef type that is available for general use.

Function in-lining A new qualifier, inline, can be applied to some C functions. A call to any function, thus qualified, will not generate a call-return sequence, but will be replaced by the assembly code associated with the function body. This expansion will save stack usage and may reduce code size if the function body is small. The assembler has always had the ability to inline small assembly sequences, so code size reductions may not be large. The operation of this qualifier is similar to use of the new #pragma inline directive. The previous (version 1.00 and earlier) inline pragma implementation has been renamed to instrinsic. (See Migration Issues below.)

Assembly psect flags To support function inlining, two new psect flags have been added: in-line and keep. These indicate to the assembler, respectively, that the contents of a psect may be inlined (and then removed), and that the contents of a psect may be inlined but must never be removed.

3.2. Version 1.01Enhanced PIC optimizations New optimizations, specifically aimed at the enhanced mid-range

PIC devices, have been added. These optimizations reduce code size and target any code that indirectly accesses locations via the FSR registers.

3.3. Version 1.00


Psect merging and splitting Two new PSECT directive flags have been added to allow splitting or merging of psects by the assembler optimizer. Now, by default, no splitting or merging takes place, but the use of the split=allow and merge=allow flags can indicate that these optimizations can take place. See the assembly language chapter in the user's guide.

Unified 8-bit device support This compiler unifies the two former HI-TECH C compilers which previously supported Microchip 8-bit PIC devices: HI-TECH C Compiler for PIC10/12/16 MCUs and HI-TECH C Compiler for PIC18 MCUs. This MPLAB XC8 compiler supports compilation for any supported 8-bit PIC device from the one application. A single device driver, xc8, is used to invoke the compiler regardless of the target device. This driver will invoke the appropriate internal applications based on your device selection. The picc and picc18 drivers which controlled the former compilers are currently still employed, and the baseline/mid-range and PIC18 code generator and assembler applications are still separate. Only one copy of the generic applications, such as the preprocessor (cpp), the parser (p1), the linker (hlink), and utilities like hexmate, objtohex and cromwell are included with the compiler, and these are shared by all device targets.

Operating modes The former HI-TECH C Compiler for PIC18 MCUs only operated in a Lite or PRO mode. With XC8, a Free (previously called Lite), Standard and PRO mode are avail-able for all target devices.

4. Migration IssuesThe following are features that are now handled differently by the compiler. These changes may require modification to your source code if porting code to this compiler version. The version number in the subheadings indicates the first compiler version to support the changes that follow.

4.1. Version 1.10Re-activation of the compiler As of version 1.10, the license files used by the XC compiler li-

cense manager are installed in a different location. This is automatically handled by the installer; however, if you plan to use an older license file with v1.10, you will need to re-activate the v1.10 XC8 compiler with your activation key before you can use it. Previously, the installers could find older license files and did not require re-activation. Re-activation will not be necessary in future when updating compiler versions.

PIC10/12/16 Assembly Header Files The names of some of the assembly header files have changed. Provided code included the generic and recommended <xc.inc> assembly header file, then this will be transparent. Specifically, changes relate to using the extension .inc instead of .h for assembly-domain header files; the names of device-specific assem-bly header files now match their C-domain counterparts.

Deprecation of in-line assembly header files The PIC10/12/16 header files that were usable from assembly that was in-line with C code have been remove. These header files were previously included via the generic file <caspic.h>. The content of these files will be in-cluded once you include <xc.h> in your C source code.


Removal of single-letter bit definitions The definitions for SFR bits that used a single letter identifier have been removed, for example the SSP1CON register bit S. These register bits are still available in the corresponding SFR structures, for example SSP1CONbits.S. The absence of these very poorly chosen identifiers should not be missed.

4.2. Version 1.01Missing SFR definitions There have been changes to the header files to ensure correlation with

the data sheet. This may resulted in some SFR definitions being no longer available. If you see undefined symbol errors or other build errors relating to SFR names, please refer to the appropriate device-specific header file and update your code.

The inline pragma The much misused inline pragma has been changed. What was the in-line pragma is now known as the intrinsic pragma. Since the inline pragma was not intended to be used with user-defined routines, this should have no impact on existing pro-jects. Use of the inline pragma in this compiler version will be ignored; however, future versions may implement a user-operable inlining feature.

Filling unused values The --FILL option, which is used to fill unused memory locations, now assumes the value specified is a 2 byte word. Previously the size of this value was the same as the target device program memory addressability i.e. 1 for PIC10/12/16 devices and 2 for PIC18 devices. The width of the value can always be specified with the option. See the user's guide section relating to hexmate.

4.3. Version 1.00Library functions removed The following library functions and macros are now longer sup-

ported and have been removed from the libraries and header files.• device_id_read

• idloc_read

• idloc_write

• config_read

• config_write

• checksum8 and the macro CHECKSUM8• checksum16 and the macro CHECKSUM16

• checksum32 and the macro CHECKSUM32

Any attempt to use these will result in a compiler error.Memory functions replaced by peripheral library equivalents Flash and EEPROM functions

and macros have been removed from the compiler libraries and header files. They have been replaced with those in the peripheral library.

Compiler drivers While you may continue to call the picc or picc18 command-line drivers, it is recommended that projects be swapped to use the unified xc8 driver. Future compiler releases may discontinue the device-specific drivers or additional functionality may be added to xc8 that will not be usable if you continue to use the device-specific drivers.


FN-type directives All FN-type directives are no longer supported and should not be used. Such directive include: FNBREAK, FNSIZE, FNROOT etc. The FNCALL and FNROOT directives are still issued by the compiler, but it is recommended that these not be used in handwritten assembly code.

5. Fixed IssuesThe following are corrections that have been made to the compiler. These may fix bugs in the generated code or alter the operation of the compiler to that which was intended or specified by the user's guide. The version number in the subheadings indicates the first compiler version to contain fixes for the issues that follow. The bracketed label(s) in the title are that issue's identifi-cation in the tracking database. This may be of benefit if you need to contact support.

5.1. Version 1.10Using far with Functions (XC8-337) If the far qualifier was used with functions, it confused

the compiler into thinking the function identifier was that of a variable destined for RAM. This may have triggered an error, if no far memory was defined, or caused a runtime code failure.

Non-functional --ADDRQUAL option (XC8-293) The reject suboption of this option was not working correctly for PIC10/12/16 device. The require and request suboptions were not working correctly for PIC18 devices. This option should now work as expected for all devices.

Conversion of integer to bit-field (XC8-353) The compiler was treating single-bit bit-fields as a boolean rather than an integer quantity when it came to assignment. Code that assigned a byte or larger integer to a single-bit bit-field may have failed.

Can't Generate Code for Duplicated Functions (XC8-358) The compiler may have issued Can't Generate Code error messages in situations where a function was duplicated in inter-rupt and main-line code and this function used pointers.

Handling of incomplete types (XC8-374) The parser was producing incorrect type information where a type used in a declaration was, at that point, incomplete, e.g. if you used a struc-ture tag in a declaration, but that tag had not been defined. The code generator would sub-sequently crash if such incomplete type information was encountered. types should always be defined before they are used, but the compiler will no longer crash on such situations.

Testing of volatile bytes (XC8-388) The code that tested volatile bytes variables for (in)e-quality was using a MOVF x,f instruction. This was not in keeping with the spirit of the volatile keyword and could also play havoc with certain registers, such as the TMR0 register, on some baseline devices, such as the 10F222, which require a read delay after being written. This code sequence is no longer used and access of SFRs will be well be-haved.

Pointer assignment failure (XC8-342, XC8-343) In some circumstances, assignment of a struc-ture member address to a pointer may have resulted in an incorrect destination pointer size


being determined by the compiler. This would have resulted in subsequent pointer derefer-ences being invalid. This problem was detected only with assignments to pointer parame-ters as part of a function call, but could have potentially occurred elsewhere.

Pointer access failure with duplicated functions (XC8-377) If a function that has a pointer pa-rameter is called from main-line and interrupt code (i.e. it is duplicated by the code genera-tor), the compiler may have issued a "looping around allocGlobals", or "can't generate code" error message. In other situations, the code may have compiled, but accesses to the pointer may have been incorrect. The issue could have affected any device with interrupts.

Incorrect access of high program memory (XC8-371) In some circumstances, exacerbated by the use of a debugger and devices with small amounts of RAM, the internal compiler sym-bol btemp may not have been defined and an error results. This symbol is now correctly defined.

Incorrect access of high program memory (XC8-363) The compiler was not correctly setting the TBLPTRU register for accesses of absolute-addressed const data located above 0xFFFF in PIC18 program memory.

Incorrect configuration bit/user ID settings (XC8-385) Attempts to program the configuration bit or user ID settings may have failed due to a sorting bug that may have resulted in the bits being programmed in the wrong order. The issue was inconsistent, but could affect all devices.

Improved error relating to IDLOC (XC8-384) If non-hex digits were used in for any nibble in the IDLOC() macro a confusing error was issued. A new message (#1436) has been created for this situation.

5.2. Version 1.01Looping around allocGlobals error (XC8-318) This error may have been triggered, but is not

specific to any particular code sequence. The compiler has been updated to ensure this trigger will not result in this error.

Access of char arrays (XC8-304) If the index expression used with an array of char consists of a non-constant expression (e.g. a plain variable) from which is subtracted a constant (e.g. myArray[idx-1]), then the index calculation may have been incorrect and the wrong element accessed. This issue mostly affected PIC18 devices and arrays no larger than a bank. It may have affected enhanced mid-range parts with array sizes larger than 256 bytes. The issue has been resolved for all devices.

Code jumps to wrong location (XC8-295) If the assembler optimizer is enabled, assembly code that directly wrote to the program counter may have been abstracted which may have caused a jump to the wrong location and code failure. C code affected would produce a lookup table of some description, but it unlikely that C code would trigger this issue. Hand-written assembly would only be affected if the option to optimize assembly source files was enabled.


Duplicated SFRs (XC8-319) Some SFR bit-field structures had duplicate entries in the device-specific header files and have been removed.

Bit objects on baseline devices Uninitialized global bit objects may not have been zeroed for baseline devices, and in some instances, the generated startup code may have corrupted other memory locations or caused the device to restart.

Call graph inconsistencies The call depth of some functions was not correctly indicated in the call graph shown in the assembly list file. The total amount of auto, parameter and tempo-rary variable usage was also incorrectly displayed for some functions in the call graph.

Wide bit-fields in header files (XC8-289) Some header files were being generated with SFR bit-fields that were wider than a byte. This is not allowed by the compiler. Such definitions are no longer contained in the header files.

Wrong external memory access (XC8-286) The --EMI options was not correctly being proc-essed and this may have meant that access to external memory may have been incorrect. Only several PIC18 devices have such memory.

Error on assembly directives after including header file (XC8-285) After including the xc8.inc header file into assembly code, some assembly directives, e.g. the DS directive, may have generated an error. This was due to SFR names conflicting with the directive's name. The header files will on longer use any SFR name that conflicts with a directive. This will mean that SFR names may not always match those listed in the device data sheet.

Bad access at start of function (XC8-94) The code at the beginning of function could have ac-cessed the wrong address (bank) for a variable if the compiled stack was built up over mul-tiple banks and WREG was used to pass a function argument.

Incorrect RAM ranges with 12F1501 (XC8-274) The RAM ranges associated with this device did not reserve the memory from 50-6F. Allocation of user-defined objects to this memory may have caused code failure.

Bad results with 24-bit expressions (XC8-227) The code generated for some arithmetic and bitwise operations involving 24-bit integers were missing some necessary bank selections which may have generated wrong results.

Compiler crash with conditional operator (XC8-117) Assigning a pointer the result of an ex-pression using nested conditional statements could have caused the compiler to crash for PIC18 devices.

Bad results with long arithmetic (XC8-241) The results of some long arithmetic expressions may not be correctly assigned due to a bank selection bug.

Build errors in MPLAB X IDE (XC8-101, XC8-104) Source file paths that contained spaces were being written to the dependency files without the spaces escaped. This resulted in these files containing erroneous targets and dependancies which may have resulted in in-correct builds. In addition, the preprocessor would generating a dependency file for assem-bly modules incorrectly assuming that the intermediate file to have a ".p1" extension and not ".obj" resulting in the same behavior.


Crash with compiler-domain symbols (XC8-18) User-define variables and functions should never start with an underscore character as such symbols are in the compiler's domain. In-appropriate use of these symbols in a program, however, was leading to a compiler crash. The crash no longer will occur; however, you should continue to avoid using symbols be-ginning with an underscore.

MPLAB X IDE plugin Previously the XC8 MPLAB IDE plugin overwrote the HI-TECH Uni-versal Plugin such that the universal plugin no longer appeared in the toolsuite list. This is no longer the case. The XC8 v1.01 installer will install and reinstate both the XC8 and uni-versal plugins so that they will both be selectable from the IDE.

Bad if() code (XC8-58) In some instances the assembler optimizer would incorrectly move MOVLW instructions which could cause some logical expressions (such as those used by if statements) to be incorrectly evaluated.

Not honoring message disable (XC8-62) When compiling for PIC10/12/16 targets the assem-bler was not honoring the --MSGDISABLE option. Thus, it was impossible to disable warn-ing messages produced by this application. This issue did not affect other applications or any application when compiling for PIC18 targets.

Bad call to subtype() Error (XC8-73) In some instances where a function has a pointer type as its first parameter but the function is never called, a "bad call to typeSub()" error may oc-cur.

Incorrect optimizations involving carry bit (XC8-77) The assembly optimizer may have incor-rectly moved instructions that set the carry bit above other code that would subsequently clear carry before it was being used.

Bad optimization of indirect access (XC8-95) In some situations, expressions with the form *ptr = *ptr op A; may fail. In particular it is the optimization of assignment-operations when the destination is an indirect access which can lead to incorrect results.

Bad right shift code (XC8-105) In some instances for PIC18 targets, the code generated for right-shift a signed long operand would access the wrong file registers, giving an incor-rect result.

Memory reservation using --CODEOFFSET (XC8-230) This option should have reserved memory in all linker classes associated with program memory. It was only reserving this memory from the CODE class but not for other classes that hold const objects. This was unlikely to cause issues since const objects are typically not allocated to low address; how-ever, all classes are now adjusted.

Badly optimized if() code (XC8-75) The assembler optimizer may have procedurally abstracted code that erroneously contained return-style instructions. This may have caused code such as if() statements to be ignored and fail. This has now been corrected.

Redundant MOVLP instructions (XC8-49) Redundant MOVLP instructions were being pro-duced for some call sequences for Enhanced mid-range devices.

5.3. Version 1.00


Bogus warning on -B option (XC8-11) If compiling under MPLAB v8, in some instances a warning indicating that the -B option was defunct was issued. This option is indeed de-funct, but the compiler has been adjusted so that this is not produced when compiling un-der the IDE.

Parser crash (PICC18-618) If an unexpected attribute was used with function definitions, e.g. __attribute__ ((const)), the parser crashed instead of giving an error. This has now been corrected.

Parser crash (PICC-684, PICC-688, PICC18-596, PICC18-607, PICC18-616, PICC18-619, PICC18-620, PICC18-621) In some circumstances, the parser would crash when encoun-tering illegal or unusual source code. This has been seen with code that accesses some structure members, passing malformed arguments to functions, or with non-prototyped (K&R) function definitions. Changes to the parser should prevent the application crashing.

Bad code associated with negation (PICC-652) In certain cases, negating a 16-bit integer may have overwritten WREG. This has been corrected.

Crash on structure with no members (PICC-597) If a structure was defined that had no mem-bers, the parser application crashed. This has been rectified.

Arrays of structures in MPLAB not watchable (PICC-544, PICC18-593) In some circum-stances, the watch window in MPLAB IDE v8 would show elements of a structure array correctly. Changes have been made to the compiler output to correct this.

Redirecting function using psect pragma (PICC-514) There were instances where use of the #pragma psect would not appear to work as the name of the psect being redirected was changing as a result of psect merging by the assembler optimizer. A new system of identi-fying mergable and splittable psects was introduced, which will assist in reducing this problem.

MPLAB IDE popup about source code using #include (PICC18-565, PICC18-566) If you are using any of the printf family of functions, an MPLAB IDE popup may have appeared warning that "The project contains source files that use the #include directive from within a function to include source code". Although this would not have affected normal debugging operations in this particular instance, the trigger for message has been adjusted so that this message will no longer be produced.

Zeroing of config bits (PICC18-600) If you used the PIC18 pragma to program only the lower byte of a configuration location, the compiler may have zeroed the upper byte rather than use the default configuration value. Ideally, all configuration words should be specified in your programs or the --RUNTIME option to program the device with the default config words should be enabled. However, this zeroing of he upper byte will no longer occur.

Undefined symbol with empty loops (PICC18-524) The compiler an remove some condition code if the state of variables used in the controlling expressions are know. In some in-stances, where the true or false statements were an empty loop (e.g. while), the compiler may have deleted a label that was still being referenced by other code. This would have produced an undefined symbol, which might look something like l9.


Crash with undefined functions (PICC18-532) If function that were not defined were assigned as the target of a function pointer, the compiler may have crashed. This crash has been fixed and now the linker will report the symbols as being undefined, as expected.

Errors indicated in unused files (PICC-428) When an error indicating a variable is too large is emitted, the name of the file that contained the variable definition may have indicated a source file that was not in the project, such as a library file.

Assignment to structure ignored (PICC-433) In situations where an assignment is made to a structure is followed by an assignment to a bitfield within that structure, the initial assign-ment may be removed. This would only occur in PRO mode. This has now been fixed and the compiler notes the distinction between the objects accessed.

Compiler crash and old library modules (PICC-573) The compiler may have crashed is some library modules were used from the SOURCES directory of the compiler. This directory no longer include files from non-OCG compilers, as they are not compatible to OCG compil-ers and will not compile.

Undefined ?Fake symbol (PICC-589, PICC-581) This error may have occurred for code that indirectly called a external function defined in assembly code or was part of another build. Only functions that returned a value in memory would be affected by this issue. The error is no longer produce, but you are required to define a symbol that represents the memory location where the return value will be stored. See the User's Guide on External Functions for more information.

Linker crash with assembler optimizers (PICC-648) If the assembler optimizers (psect merg-ing) were enabled, in some situations the linker would crash if a psect was removed but was still referenced by a directive. This has been corrected and all assembly optimizations may be employed.

Compiler crash with bad memory ranges (PICC-608) If a memory range was specified using the --RAM option, or its equivalent in MPLAB IDE, and that range was excessively large, the compiler my have produced an assertion failure. In such situations, the compiler now prints a more informative error.

Cromwell error with MCPxxxx targets (PICC-642) If compiling for an MCP device, an error relating to the "prefix list" would have been triggered. This has been corrected and compi-lation can be performed for these devices.

Cromwell crash (PICC-493) Cromwell may have crashed if encountering information gener-ated from static bit objects that are absolute.

Bad RAM ranges (PICC-608) If specifying additional memory to the --RAM option which were larger than possibly supported by the device, the driver may have generated an assertion. An error is now emitted in such situations.

Crash with malformed warning pragma (PICC-610) If the message number(s) were omitted from the #pragma warning disable directive, the compiler crashed. An error is now emitted if this situation is encountered.


Read of write-only registers (PICC-658) If an attempt was made to read a Baseline device write-only register, e.g. TRIS, the compiler would produce a can't generate code error mes-sage. A more specific message is now produced alerting you to the offending operation.

Can't generate code (PICC-683, PICC-499) The compiler was not able to generate code for some expressions involving logical or bitwise operation on bit objects. This has been cor-rected.

Prototype for eeprom_write (PICC-675) The prototype for eeprom_write incorrectly stated that this function returned a value. This was not correct and the prototype now indicates a return type of void.

CLRWDT appearing in _delay output (PICC-460) For some delay values, the _delay in-line function for PIC18 targets only produced CLRWDT instructions as part of the delay. This has been corrected and a NOP is used instead. For PIC18 targets, you have a choice of _delay or _delaywdt to implement in-line delays, which do not use and use, respectively, the CLRWDT instruction as part of the delay.

Optimization of volatile bits (PICC-606) The compiler was not generating the appropriate in-formation to prevent the assembler optimizer from optimizing access to volatile bit objects. Although the code would have been functionally correct, subsequent optimizations may have caused runtime failures. This correct information is now generated and volatile bit access will not be optimized unexpectedly.

vprintf and vsprintf linker errors (PICC-524, PICC-619) Trying to use either of these func-tions may have resulted in the undefined symbols __doprint, _vprintf or _vsprintf. These routines are now fully supported in the libraries and can be called.

Upper bound of mktime (PICC18-126) The mktime function had a upper limit of the year 2020; higher values would return incorrect results. This limit has now been extended to the year 2038.

Bad call to subtype with regsused pragma (PICC18-392) If using the regsused pragma for a function before the definition of that function, this error occurred. This has been corrected and the pragma can be used at any point in the source code.

Looping around allocGlobals error (PICC-570) This error may have occurred in some situa-tions, particularly when NULL pointers were being assigned in the source code. This has been corrected.

Read of wrong array element (PICC-542) When reading a const array using a constant index, in some situations, the wrong element may be read. This bug did not affect array access using a variable or other expression as the index.

Memory allocation An improved memory allocation scheme is used when variables are quali-fied as being in a particular bank or near and the --ADDRQUAL option is set to require. This will reduce the likelihood of Can't find space errors from the linker.

Code generator crash with undefined functions (PICC18-532) If a function pointer was as-signed the address of a function that is defined in external code, the code generator may


have crashed. External functions might include those defined in assembly source code. This only affect PIC18 targets.

Can't generate code errors with nested structures (PICC-628) This error may have occurred for code that accessed bit-fields within structures that themselves were members of other structures.

No stack allocated to function error (PICC-611, PICC-485, PICC-416, PICC-637) This error is now unlikely to be triggered by code. It was most likely to occur with complex expres-sions, particularly involving floating-point arithmetic.

Can't generate code errors with short long types (PICC-632) Some assignment operations on short long types, e.g. |=, &= and *= etc may have triggered a "can't generate code" error when compiling for baseline or mid-range parts. The ability to generate code for such ex-pressions has been enhanced.

Parser crash with bad identifiers (PICC-622) The parser (p1) may have crashed if certain keywords were used as a variable identifier, e.g. if you were to try to define a variable with the name eeprom. A crash may also have occurred if reserved keywords were used in place of variables in expressions. The crash has been corrected, but variable identifiers cannot be reserved keywords.

Symbol defined more than once for in-line assembly (PICC-563) If the REPT macro was specified in in-line assembly, the code generator was generating additional labels multiple times. This triggered an error indicating the symbol was defined more than once. This has been corrected and REPT can safely be used in in-line assembly code.

RAM ranges for 16(L)F1507 The RAM ranges available when selecting this device extended further than the physical device implementation. This has been corrected.

6. Known Issues

The following are limitations in the compiler's operation. These may be general coding restric-tions, or deviations from information contained in the user's manual. The bracketed label(s) in the title are that issue's identification in the tracking database. This may be of benefit if you need to contact support.XCLM licensing (XCLM-44) Internet activation may fail on some Linux distributions, most

notably Fedora. The error message will be "Invalid hostid type". If you see this error when activating, please enter an engineering support ticket at Support and they will be able to provide alternative activation instructions.

Installer execution On both Mac OS X and Linux, it is necessary to run the installer as root or with superuser privileges (using sudo, for example). If the installer is started without supe-ruser privileges on Mac OS X, it will exit and display an informative message. In the same situation on Linux, the installer will fail when it attempts to write to directories for which it


http://support.microchip.com/

http://support.microchip.com/

does not have adequate rights. The messages displayed will relate to these access failures. For correct operation, run the installer via sudo, or as the root user, on these systems.

PATH environment variable On Linux systems, the installer, by default, updates the PATH envi-ronment variable to include paths to the new executables being installed. If the installer is run via sudo, the default action will update the PATH variable of the user executing the sudo command. If the installer is run by root, the installer will only update root's PATH variable, and not the PATH variables of ordinary users. If installing the compiler while logged in as root, a better choice is to update all user PATH variables. Alternatively, skip the step to update the PATH variable in the installer, and to update the PATH variables of users who will use the software, manually.

C18 code compatibility The C18 code compatibility feature is only a beta implementation. Leg-acy C18 projects may not compile, or compile with runtime errors when using XC8.

Changing toolsuite in MPLAB IDE v8 If the toolsuite associated with an MPLAB IDE v8 pro-ject is changed from the HI-TECH Universal toolsuite to the XC8 toolsuite, or vice versa, the selected options in the original toolsuite are not copied over to the new toolsuite. Please note all options before you switch and confirm the options in the new toolsuite.

IDE plugin overwrite (XC8-360, XC8-367) If an older compiler is installed after installing a newer compiler, the new IDE plugin may be overwritten. (The 1.01 compiler installer is known to overwrite the 1.10 plugin.) You can manually reinstall the new plugin without having to reinstall the compiler. Run the command:regsvr32 "C:\Path_to\microchip\xc8\v1.10\bin\MPLABXC8.dll"

for example, replacing the path with that appropriate for your compiler installation.Call to empty function after comparison (XC8-398) Code which calls a function containing no

statements (an empty definition) which is preceded by an if() statement may fail for 16F1xxx parts. If the assembler optimizer is enabled, the call to the empty function, or sub-sequent code may fail. Adding any code, such as NOP();, to the empty function is enough to work around this issue.

Assignment to pointer member of a union (XC8-341) For PIC18 devices and when the com-piler is operating in PRO mode, assignment of an address to a pointer that is a member of a union which in tern is part of a structure may fail if the member is accessed via structure pointer a in the following example.strPtr->myStruct.myUnion.myPtr = mySrcPtr;

PIC12F529T39A/T48A memory restrictions The previous limitation which restricted memory to the first 4 RAM banks for user-defined variables has been lifted. Note, however, that the compiler will not allow you to define objects that span multiple banks on these devices.

In-line delay values The delay value specified with the in-line delay function (_delay) is lim-ited to 179,200 on PIC18 devices. With PIC10/12/16 devices, this value may be up to 50,659,000.


In-line delays On PIC10/12/16 targets, the only in-line delay function implemented is _delay. This uses NOP instructions, when required, as part of the delay sequence. There is currently no equivalent version of this function that uses the CLRWDT instruction. The function, _de-laywdt which is available for PIC18 targets, does use the CLRWDT instruction.

Watching const values in MPLAB v8 MPLAB IDE v8 cannot correctly display in the Watch view any object qualified as const and stored as RETLW instructions in memory. Any const object compiled for a PIC18 device will be observable, and this issue does not affect MPLAB X IDE.

Use of hardware multiply The PIC18 hardware multiply instruction is not used for any multiply operation whose operands are larger than 8 bits. A library multiply routine is used for such operations.

Disabling function duplication (XC8-3) The interrupt_level pragma current does not pre-vent duplication of functions called from both main-line and interrupt code.

Psect pragma and data psects As described in the manual, the #pragma psect directive should not be used to move initialized variables that would normally be located in one of the 'data' psects. The initial values in program memory and space for the variables them-selves in RAM must be built up in a strict order. Using this pragma will violate this as-sumption.

In-line assembly and labels Functions which are called from both main-line and interrupt code should not contain in-line assembly that defines assembly labels. Such labels will not be assigned the usual duplication prefix (i1, i2 etc) and will result in multiply-defined sym-bol errors.

Bank qualifiers Only bankx qualifiers for data banks 0 through 3 are supported by the compiler. (These are enabled using the --ADDRQUAL option). Use absolute variables to place objects in other banks, if required.

Memory Report for Absolute Variables If the user defines absolute variables (variables placed at an absolute address via the @ construct), and any of these variables overlap other vari-ables in memory, the compiler will also count overlapping variables when it comes to re-porting on memory usage. That is, if two, byte-sized variables are located at the same ad-dress, the report will indicate 2 bytes being used, rather than 1.

Out of bound variables in watch view If variables are defined for Enhanced Mid-range devices that are large and placed in the linear memory space, MPLAB IDE shows these objects are being "out of bound". This is display issue and is not indicative of code failure.

Variables shown in wrong memory space If an object is made absolute and allocated address 0, it may only be visible in the program memory view of the IDE. This is purely an IDE view issue and does not impact on the code operation.

Wrong watch values The IDE expects debug information to be available in COFF. This file format does not allow the compiler to specify the necessary information for the IDE to be able to correctly interpret some pointer variables. The compiler allocates pointer sizes based on their usage, whereas the IDE assumes all pointers are a fixed size. This can make


this issue seem "hit and miss". Any attempt to examine pointers in the IDE's Watch view may result in incorrect values being shown, or an "invalid" message being displayed in the view. These issues will correct themselves when ELF is employed to convey debug infor-mation in future IDE/compiler versions.

Switch strategies There is only one possible switch strategy currently available for PIC18 de-vices. It uses the space switch type. New strategies will be introduced in future compiler versions so that PIC18 devices have similar options to the baseline/mid-range devices.

Copying compiler header files The header files shipped with the compiler are specific to that compiler version. Future compiler versions may ship with modified header files. If you copy compiler header files into your project, particularly if you modify these files, be aware that they may not be compatible with future versions of the compiler.

SFRs for 12LF1840T48A The following are inconsistent in this device's header file.• DACNSS bit not defined in DACCON0 register (118h)

• SSP1M0, SSP1M1, SSP1M2, SSP1M3, SSPEN, SSP1OV defined as SSPM0, SSPM1, SSPM2, SSPM3, SSPEN, SSPOV in SSP1CON1 register (215h)

SFRs for 16(L)F1507 The following are inconsistent in this device's header file.

• In NCO1ACCU (49Ah) NCO1ACC<23:20> not defined

• CWG1ASD register (695h) defined as CWG1CON2

Initialization of eeprom arrays Initialization of large arrays that are qualified eeprom may fail on enhanced mid-range devices. When the size of these arrays exceeds the size of the free space in each GP RAM bank, the compiler tries to allocate the array to the linear RAM space, not EEPROM. Small arrays are not affected by this issue.

Initialization of auto structures Structures which are auto cannot be initialized. Simply define the structure and assign values separately.

EEPROM variable limitations There are some limitations to using variables qualified as ee-prom. Avoid using these variables in complicated expressions and remember that code which accesses such objects is complex and very slow.

Size of EEPROM arrays Arrays that qualified as eeprom and which are initialized may not have all initial values written to the EEPROM.

Constant index into int array If a constant is used as the index into an array of const int type, the element read may be incorrect. This issue does not affect arrays of different types, or when the index is a variable.

Absolute initialized variables Variables which are absolute and which are not const cannot be initialized. The following example will not generate an error, but will not work as ex-pected. Define the variable as absolute and initialize it in main-line code.unsigned int varname @ 0x0200=0x40; // will not work as expected

Stack overflow When the managed stack is used (the stackcall suboption to the --RUNTIME option is enabled) in some situations the stack may overflow leading to code failure. With


this option enabled, if a function call would normally overflow the stack, the compiler will automatically swap to using a lookup table method of calling the function to avoid the overflow. However, if these functions are indirect function calls (made via a pointer) the compiler will actually encode them using a regular call instruction and when these calls return, the stack will overflow. The managed stack works as expected for all direct function calls, and for all indirect calls that do not exceed the stack depth.

Main function size If the function main produces assembly code that is 0x1FF words long, this cannot be placed in the program memory and a "can't find space" error message is pro-duced. Increasing or decreasing the size of the function will allow it to be positioned cor-rectly and the error will not be displayed. This only affects Baseline PIC devices.

Functions called from in-line assembly The code generator will not be able to identify a C function called only from in-line assembly code if the definition for that C function is placed before the assembly call instruction in the source file. Placing the function defini-tion after the call is acceptable. If the function cannot be identified, no code will be gener-ated for the function and the linker will issue undefined symbol errors.

Can't Generate Code messages When compiling for baseline devices, some complex expres-sions may cause compile-time errors (712) Can't generate code for this expression. The ex-pressions should be simplified to work around this. This may require the use of additional variables to store intermediate results. This is most likely with long integer or floating-point arithmetic and particularly those devices with less than 4 bytes of common memory available.

Indirect function calls The parameters to functions called indirectly via a pointer may not be passed correctly if the following conditions are met.

• A function is called indirectly from both main-line and interrupt code

• This function has parameters

• The function pointer used to call this function is also assigned the address of other func-tions at some point in the program; and

• This function pointer is initialized when it is defined.All other instances of indirectly calling a function, even from interrupt and main-line code, will work as expected.

option and tris For baseline devices, the OPTION and TRIS registers must be written as a byte. Writing individual bits is not supported.

PIC17 support PIC 17 devices (for example, 17C756) is not supported by this compiler.fast32 floats The option to select the fast 32-bit float or double library for PIC17 devices that

was included in the PICC STD compiler is no longer available.Procedural abstraction An instance has been seen, but not resolved, which appears to be related

to the procedural abstraction optimization of the assembler. If suspect behavior is encoun-tered, compiling for speed, instead of space, will disable this optimization and allow you to


confirm if this is the underlying cause. This problem was reported for an Enhanced Mid-range device.

Pointer assignments In PRO mode, if a pointer assignment is made where one of the pointers is a member of a structure, the compiler may fail to copy the whole pointer, causing the pointer to register an incorrect address.

Configuration words (PIC18 parts only) The new device support introduced in PICC18 v9.80 will not automatically program the default values into the configuration words when no value is specified. If your project does not program all configuration words explicitly, se-lect the option "Program the device with default config words" in the Linker tab.

Specifying configuration words on PIC10/12/16 devices The __PROG_CONFIG() and __CON-FIG() macros can be used to specify the configuration words on PIC10/12/16 devices as well as PIC18 devices. The __PROG_CONFIG() macro must use a literal constant argu-ment; you cannot use the configuration setting symbols with this macro. The __CONFIG() macro must only use the predefined configuration setting symbols and you may not not use a literal value with this macro.

MPLAB 8 debugging information When compiling projects with multiple C files, MPLAB may throw up a warning dialog with the message beginning "The project contains source files that use the #include directive from within a function to include source code." This is due to the Omniscient Code Generation in the XC8 compiler, which can make multiple source files appear in the one COF output. It does not affect the operation of the actual pro-ject.

Selecting compiler mode When using the Free version of the XC8 compiler, the ability to select compilation modes in the plugin is disabled. If you switch to an older version of the com-piler inside the MPLAB 8 plugin, you will not be able to change the compilation mode to Standard or PRO. To work around this behavior, close the plugin, then reopen it. This should allow other compilation modes to be selected.

rfPIC12 parts To use the rfPIC12 parts, for example the rfPIC12C509AF, you will need to spec-ify to the compiler a part name in a format similar to RF509AF, for example. You can also use an alias like 12C509AF, for example. The full part name is also not appropriate when compiling from MPLAB IDE.

MPLAB IDE integration If Compiler is to be used from MPLAB IDE, then you must install MPLAB IDE prior to installing Compiler.

7. Microchip Errata

For PIC18 devices, this release of the XC8 compiler recognizes the published silicon errata is-sues listed in the table below. Some of these issues have been corrected and no longer apply in recent silicon revisions. Refer to Microchip's device errata documents for details on which issues are still pertinent for your silicon revision. The compiler's chip configuration file records which


issues are applicable to each device. Specific errata workarounds can be selectively enabled or disabled via the driver's --ERRATA command line option.

Name Description Workaround details

4000 Execution of some flow control operations may yield unexpected results when instructions vector code execution across the 4000h address boundary.

Each block of program code is not allowed to grow over the 4000h address boundary. Additional NOP instructions are inserted at prescribed locations.

LFSR Using the LFSR instruction to load a value into a specified FSR register may also corrupt a RAM loca-tion.

The compiler will load FSR registers without us-ing the LFSR instruction.

DAW The DAW instruction may improperly clear the CARRY bit (STATUS<0>) when executed.

The compiler is not af-fected by this issue.

MINUS40 Table read operations above the user program space (>1FFFFFh) may yield erroneous results at the extreme low end of the device’s rated temperature range (-40o C).

Affected library sources config.c, idloc.c and de- vread.c employ additional NOP instructions at prescribed locations.

EEDAT When reading EEPROM, the contents of the EE-DATA register may become corrupted in the second in-struction cycle after setting the RD bit (EECON1<0>).

The EEPROM_READ macro and eeprom_read library function read EEDATA im-mediately.



EEADR The result returned from an EEPROM read operation can be corrupted if the RD bit is set immediately fol-lowing the loading of the EEADR register.

The compiler is not af-fected by this issue.

EELVD Writes to EEPROM mem-ory may not succeed if the internal voltage reference is not set.

No workaround applied

FLLVD Writes to program memory may not succeed if the in-ternal voltage reference is not set


RESET A GOTO instruction placed at the reset vector may not execute.

Additional NOP instruction inserted at reset vector if following instruction is GOTO

FASTINTS If a high-priority interrupt occurs during a two-cycle instruction which modifies WREG, BSR or STATUS, the fast- interrupt return mechanism (via shadow registers) will restore the value held by the register before the instruction.

Additional code reloads the shadow registers with the correct values of WREG, STATUS and BSR.

BSR15 Peripheral flags may be er-roneously affected if the BSR register holds the value 15, and an instruction is executed that holds the value C9h in its 8 least sig-nificant bits.

Compiler avoids generating MOVLB 15 instructions. A warning is issued if this instruction is detected.



TBLWTINT If a peripheral interrupt oc-curs during a TBLWT op-eration, data can be cor-rupted.

Library routine flash_write() will temporar-ily disable all applicable interrupt-enable bits before execution of a TBLWT in-struction.

FW4000 Self write operations initi-ated from and acting upon a range within the same side of the 4000h boundary may fail based on se-quences of instructions executed following the write.


RESETRAM Data in a RAM location can become corrupted if an asynchronous reset (e.g. WDT, MCLR event) oc-curs during a write opera-tion to that location.

A warning will be issued if the length nvram psect is greater than zero bytes (persistent variables popu-late this psect).


MPLAB XC8 C Compiler Version 1.10 Release...

Documents

Transcript of MPLAB XC8 C Compiler Version 1.10 Release...